Beautifying Prometheus metrics using Grafana

on

     

Beautifying Prometheus metrics using Grafana

This is a buildup on my previous articles
I have always maintained that Prometheus has limited metrics visualization. It has good enough visualization that leaves you wanting for more.
Grafana covers visualization well. And on top of that - adds alerting capabilities too.

In our company, we use the Prometheus + Grafana stack to monitor the Kubernetes cluster, Spring Boot applications, Infrastructure applications like Gitlab, Jenkins, etc.

What is covered

This article covers
  • installing Grafana as a Docker container
  • connecting to Prometheus
  • using an existing dashboard
  • creating a new dashboard to monitor endpoints

Installing Grafana as a Docker container

docker run --detach \
  --publish 3000:3000 \
  --name grafana \
  grafana/grafana

Nothing fancy here. But if you wish to go fancy - then here is the documentation.

Open http://localhost:3000/login to login. Use admin/admin to login. That is the default id-password. 

My home page looks like this:

Connecting to Prometheus

  • On the home page, there is a box "Data Sources". Click it.
  • Select Prometheus.
  • Select URL as http://docker.for.mac.host.internal:9090.
    • Note: instead of localhost, I use 'docker.for.mac.host.internal' because Grafana is running as a docker container. And to access my laptop's network, I need to use a special DNS name. Or stackoverflow.
  • Click "Save and Test" button.
My configurations look like this:

Using an existing Dashboard

Grafana has a big community. There are many pre-built dashboards that we can use. One of the most widely used is "JVM Dashboard". 

Step 1: Choose a Dashboard

  • Open https://grafana.com/grafana/dashboards. This page allows us to search among dashboards.
  • Scroll down until you find a filter panel on the left side of the page.
  • Select "Name" as "JVM, and "Data Source" as "Prometheus". Now choose the one you like.
  • I choose the one by the author "mweirauch". At the time of writing this article, there are 57822 downloads. Hence I trust this one.
  • Select the dashboard you want. It takes you to the dashboard description page. Copt its URL. Mine looks like this: https://grafana.com/grafana/dashboards/4701.



Step 2: Install selected Dashboard

  • On the Home page, on the left menubar, there is a + symbol. Hover over it and select "Import". We shall try to import an existing Dashboard created by the community.
  • Put the URL (of the selected dashboard) and click "Load"
  • You should see details about the selected dashboard. 
  • In the "Prometheus" dropdown (towards the bottom of the form), select the DataSource you had created earlier. I had named mine "Prometheus". So I select that.
  • Click "Import"

And ta-da!
I see my dashboard.

Create a new Dashboard

Someone else's Dashboard is his view of his application. For your application, you will want to create something your own.


I wish to see what's happening with my /chance endpoint (from the previous article).


  • Go to home page
  • On left menubar, hover over + symbol. Select Dashboard.
  • Click on "Add an empty panel"
  • This opens up a blank graph with a form at the bottom. Notice that the textbox needs a Prometheus Query. Use the query from the previous article 
    • http_server_requests_seconds_count{method="GET", uri="/chance"}
  • And click "Apply". This button is placed in the top-right corner.
  • Don't forget to save your dashboard - click on the save icon on the top-right of the page.


Ta-da!

My dashboard leaves me wanting more - but it is a good start.


Conclusion

Grafana spits out beautiful charts. It has alerting module too. Talking about alerts is beyond scope of this article.

I would go ahead to say that Grafana makes Prometheus more usable. 

Comments

Post a Comment